Does tailoring matter? Meta-analy...
Does Tailoring Matter? Meta-Analytic Review of Tailored Print Health Behavior Change Interventions Seth M. Noar, Christina N. Benac, and Melissa S. Harris University of Kentucky Although there is a large and growing literature on tailored print health behavior change interventions, it is currently not known if or to what extent tailoring works. The current study provides a meta-analytic review of this literature, with a primary focus on the effects of tailoring. A comprehensive search strategy yielded 57 studies that met inclusion criteria. Those studies���which contained a cumulative N 58,454���were subsequently meta-analyzed. The sample size-weighted mean effect size of the effects of tailoring on health behavior change was found to be r .074. Variables that were found to significantly moderate the effect included (a) type of comparison condition, (b) health behavior, (c) type of participant population (both type of recruitment and country of sample), (d) type of print material, (e) number of intervention contacts, (f) length of follow-up, (g) number and type of theoretical concepts tailored on, and (h) whether demographics and/or behavior were tailored on. Implications of these results are discussed and future directions for research on tailored health messages and interventions are offered. Keywords: tailored message, health communication, behavior change, theory, intervention According to Mokdad, Marks, Stroup, and Gerberding (2004, 2005), approximately half of all deaths that occur each year in the United States are preventable. This is the case because such deaths are caused by largely preventable and modifiable behavioral risk factors. For instance, it is estimated that 2,403,351 individuals died in the United States in 2000 nearly half of these (1,124,000) were due to largely modifiable factors, including the use of tobacco, poor diet and physical inactivity, alcohol consumption, microbial agents (such as influenza and pneumonia), toxic agents (e.g., air pollutants such as asbestos), motor vehicle accidents, firearm injuries, unsafe sexual behavior, and illicit drug use. Such behav- iors are major contributors to the development of leading causes of death, such as heart disease, stroke, and numerous cancers (Mok- dad et al., 2004, 2005). Moreover, the three key behaviors of tobacco use, unhealthy diet, and lack of physical activity ac- counted for approximately 71% of the more than 1 million pre- ventable deaths in the year 2000 (Mokdad et al., 2004, 2005), indicating that these three behaviors deserve unique attention. As these preventable risk factors are themselves behaviors, individuals may be able to add years to their lives as well as reduce substantial suffering if they are willing and able to make the health behavior changes necessary to potentially avoid chronic disease and premature death. Unfortunately, data on the enactment of such health behaviors in the United States are alarming. For instance, national data from the 2005 Behavioral Risk Factor Surveillance Survey indicate that among U.S. adults, only 27.4% exercise regularly (defined as 20 or more minutes of vigorous physical activity 3 or more days per week), 23.2% eat five or more servings of fruits and vegetables per day, and 79.5% are nonsmokers (Centers for Disease Control and Prevention, 2007). In addition, an analysis of Behavioral Risk Factor Surveillance Survey data from the year 2000 found that across the four key behaviors of adequate exercise, healthy diet, smoking, and maintaining a healthy weight, only 3% of U.S. adults met criteria for all four (Reeves & Rafferty, 2005). This indicates that those who engage in one health behavior may very well not engage in another. Given that trends suggest that deaths attributable to poor diet and physical inactivity increased by 22% between the years 1990 and 2000 (Mokdad et al., 2004, 2005), future trends may prove even more challenging to public health officials and those faced with the task of turning these trends around. Thus, these data, taken together, suggest that innovative and promising approaches to health behavior change are vitally needed. In fact, one of our greatest public health challenges is developing health behavior change programs and interventions to improve the health and reduce the burden of chronic disease of Americans and individuals worldwide (Glanz, Rimer, & Lewis, 2002). This is a task that health psychologists and those in related disciplines are uniquely qualified to undertake. A Tailored Message Approach to Health Behavior Change The health behavior change literature is vast and includes ap- proaches based on a number of behavioral theories as well as approaches that operate at a number of levels, including individual, interpersonal, group, and community (e.g., DiClemente, Crosby, & Kegler, 2002 Glanz et al., 2002 Institute of Medicine, 2000). A common thread that runs through all of this research has to do with Seth M. Noar, Christina N. Benac, and Melissa S. Harris, Department of Communication, University of Kentucky. Melissa S. Harris is now at the Pacific Institute for Research and Evaluation, Nashville, Tennessee. This research was supported in part by funds provided by the College of Communications and Information Studies, University of Kentucky. Correspondence concerning this article should be addressed to Seth M. Noar, Department of Communication, University of Kentucky, 248 Grehan Building, Lexington, KY 40506-0042. E-mail: email@example.com Psychological Bulletin Copyright 2007 by the American Psychological Association 2007, Vol. 133, No. 4, 673���693 0033-2909/07/$12.00 DOI: 10.1037/0033-2909.133.4.673 673
effective health communication. How can we create and deliver messages to the public that are relevant, interesting, informative, and ultimately have the greatest chance of being persuasive? One blossoming area of research that has attempted to address this question is the area of tailored health messages. Kreuter, Strecher, and Glassman (1999) have described the full range of types of health communication, from messages that are not at all individualized to those that are quite individualized. Generic com- munication is defined as communication that is not individualized or based on any kind of individual assessment. An example of this is a brochure on the risks of smoking that one might read in a doctor���s office. Personalized generic communication is virtually the same as generic communication, except that it uses a charac- teristic, such as one���s name, to personalize the message. A mass mailing from a health agency or doctor���s office might be described as personalized generic communication. Targeted communication refers to messages that are developed with a certain segment of the population in mind, and the practice of message targeting is one that has been widely applied in the health education and health communication literature (e.g., Kreuter & Wray, 2003 Rimal & Adkins, 2003). In fact, most health education materials developed and used in interventions are best described as targeted commu- nication, and this practice was adapted from advertising in which dividing consumers into market segments and targeting commu- nications to those segments is an age old practice (Grunig, 1989). Although message targeting is a staple practice within health communication interventions, recent theoretical as well as techno- logical advances have led to a blossoming literature on tailored communication (e.g., Kreuter, Farrell, Olevitch, & Brennan, 2000 Revere & Dunbar, 2001 Skinner, Campbell, Rimer, Curry, & Prochaska, 1999). Kreuter, Farrell, et al. (2000) define tailoring as ���any combination of strategies and information intended to reach one specific person, based on characteristics that are unique to that person, related to the outcome of interest, and derived from an individual assessment��� (p. 277). Thus, tailored communication is uniquely individualized to each person, whereas targeted messages are developed to be effective with an entire segment of the pop- ulation. Tailored messages, however, do require individualized assessments of members of the population to develop such com- munications. In addition, although interpersonal communication is the most individualized form of communication and is used in a variety of health education interventions (e.g., brief counseling interven- tions), the potential ability to reach large audiences through computer-based tailoring of messages gives this approach major promise. In fact, Abrams et al. (1996) have argued that although individual-level psychological approaches to health behavior change have been the most efficacious, public health approaches that consider entire populations are capable of the widest reach. Tailored health message interventions have the potential to be both efficacious and, through the use of computer-based tailoring, may reach thousands of individuals (Abrams, Mills, & Bulger, 1999 Prochaska, Velicer, Fava, Rossi, & Tsoh, 2001). Although only a small number of studies has examined the cost-effectiveness of tailored interventions (e.g., Lairson, Newmark, Rakowski, Tiro, & Vernon, 2004), the existing evidence suggests that such interven- tions may be cost-effective as well. Thus, the ultimate impact of such interventions could be quite large. Health Behavior Theory and Tailored Messages A theoretical perspective that has been a driving force in the tailored message arena is the transtheoretical model (TTM) and stages of change (Prochaska & DiClemente, 1983 Prochaska, DiClemente, & Norcross, 1992). The TTM is a health behavior change theory that posits that individuals progress through five stages of change on their way toward adopting a healthy behavior or toward cessation of an unhealthy behavior. These stages include precontemplation (not intending to change), contemplation (in- tending to change in the foreseeable future), preparation (planning to change very soon and currently taking measurable steps to change), action (changed in the past 6 months), and maintenance (changed and sustained the behavior change for 6 months or more). The TTM describes the change process as cyclical rather than linear, as individuals may move forward through stages, backslide, and then continue cycling and recycling through the stages of change. A number of factors that may help propel individuals through the stages of change include increased positive percep- tions and decreased negative perceptions of making the health behavior change (Prochaska et al., 1994), increased self-efficacy that one has the skills and abilities to make the change (Prochaska, Redding, & Evers, 2002), and a variety of cognitive and behavioral change strategies or processes of change (see Prochaska et al., 1992). The TTM suggests that because individuals��� attitudes, strate- gies, and skills differ at varying stages of the change process, interventions should be uniquely tailored to those stages. Rather than a ���one size fits all��� approach, interventions should be sensi- tive to where individuals are in the change process, and messages tailored to those stages are likely to be the most effective in moving individuals forward through the stages (Prochaska, Di- Clemente, Velicer, & Rossi, 1993 Velicer et al., 1993). Not surprising, a large number of tailored message interventions have been based upon the TTM or stages of change (e.g., see Kreuter, Farrell, et al., 2000 Revere & Dunbar, 2001 Skinner et al., 1999). It should be noted, however, that although the developers of the TTM advocate using the full model���including stages of change, decisional balance, self-efficacy, and processes of change (e.g., Prochaska et al., 2002)���a number of studies in the tailored mes- sage area utilize a stages of change model in which the stages are used as the sole theoretical perspective or in combination with other health behavior concepts or theories. Thus, in the current article we refer to both the stages of change model as well as the TTM, to distinguish these two perspectives from one another. In addition, a number of other health behavior theories have been widely used as a basis for tailoring health behavior change messages (e.g., Kreuter, Farrell, et al., 2000 Revere & Dunbar, 2001 Skinner et al., 1999). These theories all suggest a number of individual-level factors that affect behavior change, and as such lend themselves to tailoring at the individual level. For instance, the health belief model (HBM Becker, 1974 Janz & Becker, 1984) posits that a key determinant of whether an individual adopts a healthy behavior is that individual���s perceived threat of a disease or negative outcome. Perceived threat is made up of two components���susceptibility, or the perception that one is at risk for a disease, as well as severity, or the perception of the seriousness of that disease. From this perspective, a prerequisite for behavior change is an individual recognizing that he or she is at risk and that 674 NOAR, BENAC, AND HARRIS
the seriousness of the disease or outcome is severe enough to motivate protective action. In addition, the model posits that weighing perceived benefits and barriers to behavior change is also important, as those viewing more benefits than barriers are more likely to take action than those viewing more barriers than benefits. Finally, more recently HBM proponents have suggested the addi- tion of self-efficacy to the model (Rosenstock, Strecher, & Becker, 1988). Self-efficacy is defined as the situation-specific confidence that one can execute a behavior to achieve a desired outcome (Bandura, 1986). A large body of literature finds that those with higher self-efficacy are more likely to implement health behavior changes as compared with those with lower self-efficacy (Bandura, 1998 Strecher, DeVellis, Becker, & Rosenstock, 1986). In addition, the theory of reasoned action (TRA Fishbein & Ajzen, 1975) and the theory of planned behavior (TPB Ajzen & Madden, 1986) posit that the most proximal predictor of health behavior is behavioral intention, or the perceived likelihood of performing a behavior. According to the TRA, intention is influ- enced by both attitudes and subjective norms regarding the behav- ior. Thus, the more positive one���s attitude as well as the more one perceives normative pressure to engage in the behavior, the more likely it is that behavioral intentions will be strengthened and the behavior will be carried out. The TPB suggests that a third factor, namely perceived behavioral control, is an important determinant of behavioral intentions. Perceived behavioral control refers to the extent to which one believes a behavior is under one���s volitional control. From the perspective of the TPB, those with more positive attitudes, perceived normative pressure, and perceived behavioral control over the behavior are more likely to form strong behavioral intentions and to engage in the behavior itself. Finally, social cognitive theory (SCT Bandura, 1986) is a comprehensive theory of behavior change that posits that health behaviors must be understood in the context of reciprocal deter- minism, or the idea that characteristics of a person, one���s environ- ment, and the behavior itself all interact and determine whether a behavior is performed. SCT suggests, however, that the most central determinant of health behavior change is self-efficacy, a concept discussed above that is now included in numerous theories of health behavior (Noar, 2005 Noar & Zimmerman, 2005). SCT suggests that in addition to confidence in performing a behavior, an individual must also believe that engaging in the behavior will lead to desirable outcomes, which are referred to as outcome expectancies. Thus, according to this perspective, individuals are most likely to engage in a health behavior if they possess the perceived ability to perform the behavior (self-efficacy) as well as the belief that engaging in the behavior will lead to expected, desirable outcomes (outcome expectancies). Tailored Messages and Health Behavior Change: What Do We Know? Although the evidence to date suggests that tailored messages are likely to be viewed as more relevant than more generic com- munications (e.g., Kreuter et al., 1999 Kreuter & Wray, 2003), a question posed in the current meta-analysis is whether such mes- sages can result in greater health behavior change as compared with generic or targeted messages. In other words, does tailoring matter, and if so, how much does it matter? Although tailored messages may be found to be more effective, the effort that goes into creating such messages is great. Thus, the effects must be large enough to warrant the investment in tailoring technology and individualization of messages (Halder et al., 2006). If the effects are not larger than targeted communication, then the additional resources needed to create individually tailored messages might be better spent in other ways, and perhaps targeting techniques (which operate at the group level) should be used instead (Kreuter & Skinner, 2000 Kreuter & Wray, 2003 Ryan, Skinner, Farrell, & Champion, 2001). A number of narrative reviews of the tailored health communi- cation literature have, in fact, examined the issue of impact of tailored messages on health behavior change. Skinner et al. (1999) reviewed 13 health behavior intervention trials testing the efficacy of tailored print messages versus nontailored comparison or con- trol conditions. They concluded that tailored messages are indeed more effective in influencing health behavior change as compared with the other conditions tested, noting that 6 of 8 studies com- paring tailored messages to similar but nontailored messages re- sulted in significant findings. Rimer and Glassman (1999) re- viewed 17 cancer communication intervention trials testing the efficacy of tailored print communications and similarly concluded that evidence suggests behavioral outcomes are more positive than they are null or negative. Kroeze, Werkman, and Brug (2006) reviewed 30 studies on computer-tailored materials for physical activity and dietary behavior change and described the evidence supporting the effectiveness of dietary computer-tailored interven- tions as ���quite strong��� (p. 208). They also concluded that too few studies existed in the physical activity domain to draw conclusions. Revere and Dunbar (2001) reviewed 37 health behavior interven- tion trials, including those utilizing print materials, automated telephone, computers, and mobile communications. They found that 34 of the 37 trials had statistically significant or improved outcomes and thus concluded that tailored interventions are effec- tive. Other reviewers of this literature have similarly concluded or suggested that tailoring appears to ���work��� (Brug, Campbell, & van Assema, 1999 Kreuter, Farrell, et al., 2000 Strecher, 1999 Ve- licer, Prochaska, & Redding, 2006). All of these conclusions about the state of the tailored message literature, however, are derived from narrative reviews, and meta- analytic scholars have often pointed to the shortcomings of the narrative review method (e.g., Johnson, Scott-Sheldon, Snyder, Noar, & Huedo-Medina, in press Rosenthal, 1991). For instance, many narrative reviews lack systematic and thorough searches of the literature, and most rely heavily on statistical significance as the sole criterion for judging the outcomes of studies. In addition, narrative reviewers often have difficulty assessing which charac- teristics of studies are associated with stronger effects. Moreover, meta-analyses yield effect sizes that provide precise estimates regarding particular phenomena, and such estimates have proven to be quite useful in numerous areas of health communication (see Noar, 2006a). In fact, a small number of meta-analyses related to the current study have recently appeared in the literature. Shaw et al. (2005) meta-analyzed 15 tailored interventions directed toward health care professionals, although the results were largely incon- clusive. Lancaster and Stead (2007) meta-analyzed self-help ma- terials for smoking cessation and included 17 tailoring studies in their analysis. They found some evidence for the effectiveness of tailored materials, although the effect sizes were quite small. Finally, Edwards et al. (2007) conducted a meta-analysis on per- 675 META-ANALYTIC REVIEW OF TAILORED INTERVENTIONS